illusory line motion and transformational apparent motion during continuous flash suppression

Illusory line motion and transformational apparentmotion during continuous flash suppression1

YUKI YAMADA* Kyushu University

TAKAHIRO KAWABE NTT Communication Science Laboratories

Abstract: A static bar is perceived to dynamically extend from a peripheral cue(illusory line motion (ILM)) or from a part of another figure presented in the previousframe (transformational apparent motion (TAM)). We examined whether visibility forthe cue stimuli affected these transformational motions. Continuous flash suppres-sion, one kind of dynamic interocular masking, was used to reduce the visibility for thecue stimuli. Both ILM and TAM significantly occurred when the d’ for cue stimuli waszero (Experiment 1) and when the cue stimuli were presented at subthreshold levels(Experiment 2). We discuss that higher-order motion processing underlying TAM andILM can be weakly but significantly activated by invisible visual information.

Key words: consciousness, awareness, motion perception, visual illusion.

A static bar is often perceived to extend from aprecued location. This phenomenon is calledillusory line motion (ILM; Hikosaka, Miyauchi,& Shimojo, 1993a,b; Figure 1). Previous studieshave shown that ILM occurs due to a presenta-tion of visual, auditory, and somatosensory cues(Bavelier, Schneider, & Monacelli, 2002; Chica,Charras, & Lupiáñez, 2008; Ghorashi, Jefferies,Kawahara, & Watanabe, 2008; Hikosaka et al.,1993a,b; Hikosaka, Miyauchi, & Shimojo, 1996;Ishigami, Klein, & Christie, 2009; Kawahara,2002; Shimojo, Miyauchi, & Hikosaka, 1997;Yamada, Kawabe, & Miura, 2008). The cuestimuli may be necessary in order for the appar-ent motion mechanism to take part in the pro-cessing for ILM (Kawahara & Yokosawa, 2001;Kawahara, Yokosawa, Nishida, & Sato, 1996;von Grünau, Dubé, & Kwas, 1996).

A static bar is perceived to extend from theside of a preceding cue that has geometrical

properties that are concordant with the staticbar. This second kind of illusory motion isknown as “transformational apparent motion”(TAM; Tse, Cavanagh, & Nakayama, 1998).TAM occurs as a result of parsing and matchingof objects within a frame or across frames.Several geometric properties, such as theconcavity/convexity of edges (Hoffman &Richards, 1984; Singh, Seyranian, & Hoffman,1999) and the smooth arrangement of Gabororientation (Field, Hayes, & Hess, 1993),strongly contribute to the formation of visualobjects in a static scene. These properties arealso the determinants of the transformationbetween visual objects in a dynamic scene,TAM (Tse et al., 1998). A recent study demon-strated that 3-D form analysis also contributesto TAM (Tse & Logothetis, 2002); hence, TAMis considered to be one kind of high-levelmotion perception that is not simply accounted

*Correspondence concerning this article should be sent to: Yuki Yamada, Faculty of Human-EnvironmentStudies, Kyushu University, Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan. (E-mail: [email protected])

1This work was supported by a Grant-in-Aid for JSPS Fellows from the Japan Society for the Promotion ofScience (Y. Y.) and the Kyushu University Research Superstar Program (T. K.).

Japanese Psychological Research2012, Volume ••, No. ••, ••–••

© 2012 Japanese Psychological Association. Published by Blackwell Publishing Ltd.

doi: 10.1111/j.1468-5884.2012.00512.x

for by means of outputs from spatiotemporallyoriented motion detectors (Fracasso, Cara-mazza, & Melcher, 2010; Tse, 2006; Tse &Caplovitz, 2006).

Between ILM and TAM, different kinds ofprocessing likely cause a similar illusory motionof a static bar. In ILM, the cue and the trailingbar are processed in a preattentive, apparentmotion mechanism (Kawahara et al., 1996). Arecent study has shown that ILM was inducedby a cue that was subjectively invisible due toobject substitution masking (Blanco & Soto,2009), suggesting that ILM occurs without avisible cue. The results of Blanco and Soto(2009) are consistent with the finding thatapparent motion also occurs without awarenessof the previous frame (Blake,Ahström, & Alais,2000). In TAM the geometric relationshipbetween cue and the trailing bar is analyzed inhigher-level visual processing (Tse & Logoth-etis, 2002). However, no studies have made anattempt to clarify whether TAM occurs withouta visible cue.

The present study aimed at investigatingwhether ILM and TAM occurred when the vis-ibility for cue stimuli was controlled by con-tinuous flash suppression. In Blanco and Soto(2009), object-substitution masking was usedto render the cue for ILM (i.e., the ILM cue)invisible. In object-substitution masking,four-dot maskers and a maskee should havecommon onsets while the maskee should dis-appear earlier than the maskers. Hence, inobject-substitution masking it is hard to renderthe cue for TAM (i.e., the TAM cue) invisiblebecause the TAM cue should physically existduring and simultaneously disappear with thepresentation of the trailing bar. Therefore, toreduce the visibility of the TAM cue, weemployed a dynamic interocular masking tech-nique called “continuous flash suppression”(Tsuchiya & Koch, 2005) in which static (ornonsalient dynamic) stimuli exposed to oneeye are rendered invisible because of the peri-odical swapping of salient colorful squaresexposed to the other eye.

Percept

Illusory line motion

Cue

Percept

Transformational apparent motion

Cue

Stimulus Stimulus

Figure 1 Examples of stimuli and percept of illusory line motion (left) and transformational apparent motion(right). The arrow on the horizontal bars represents the perceived motion direction.

Y. Yamada and T. Kawabe2

© Japanese Psychological Association 2012.

It has been predicted that ILM would occurunder continuous flash suppression if an invis-ible cue was enough to cause ILM, as shown inBlanco and Soto (2009). In addition, we pre-dicted that TAM would also occur under con-tinuous flash suppression if the spatiotemporalanalysis of visual forms, which was necessaryfor TAM, occurred independently of the cuevisibility.

In Experiment 1, to precisely learn the rela-tion between cue visibility and ILM/TAM, wefirst measured the detection thresholds for theILM and TAM cues separately by controllingthe luminance contrast of continuous flashes.Second, using the cue stimuli above and belowthe detection thresholds, we further examinedthe effect of the visibility of the cue stimuli onILM and TAM. The visibility of the cue stimuliwas assayed by perceptual sensitivity, d’, basedon signal detection theory (Green & Swets,1966). Then in Experiment 2, we re-examinedthe effect of the visibility of the cue stimuli onILM and TAM under conditions in which thecue stimuli were presented at subthresholdlevels.

Experiment 1

Method

Observers. Ten observers including the twoauthors participated in Experiment 1. Theyreported that they had normal or corrected-to-normal visual acuity. Apart from the authors,the observers were naive as to the purpose ofthe experiment.

Apparatus. The stimuli were presented on a19-in. CRT monitor (RDF193H, Mitsubishi,Japan). The resolution of the monitor was1024 ¥ 768 pixels, and the refresh rate was100 Hz. The presentation of stimuli andcollection of data were computer-controlled(Mac Pro; Apple). Using a photometer (3298F;Yokogawa, Japan), we performed a correctionfor the luminance emitted from the monitor.Observers viewed stimuli through a mirror ste-reoscope (Screenscope; Stereoaids, Australia).

Stimuli. The stimuli were generated byMATLAB (Mathworks Inc.) with the Psych-toolbox extension (Brainard, 1997; Pelli, 1997).The stimuli consisted of two fixation crosses,two square outlines, continuous flashes, andtransformational motion figures (ILM andTAM; Figure 2) displayed on a gray back-ground (45.8 cd/m2). To ensure binocularfusion, we presented a square outline to eacheye. The square outlines had a side of 14.2 deg,and a border width of 0.2 deg.The luminance ofthe border was 66.5 cd/m2. Continuous flasheswere achromatic mesh grids (15 ¥ 9 cell) inwhich the luminance of each cell was randomlyselected from 0 to 45.8 cd/m2 at 8.3 Hz. Thecenter of the mesh grids was set at 4.5 degto the left and right of the center of a squareoutline, and at 2.6 deg above the fixation. Theside length of each cell was 0.5 deg. TheILM figure comprised a cue square (ILMcue; 0.5 ¥ 0.9 deg) and a horizontal bar(0.5 ¥ 7.6 deg), and was positioned 4.0 degabove the fixation cross. The ILM cue was pre-sented 3.3 deg to the left or right of the fixationcross; the horizontal position of the bar wasalways at the center.The TAM figure comprisedan H-shaped figure with two stalks(3.8 ¥ 0.9 deg) and a small part of a horizontalbar (TAM cue; 0.5 ¥ 0.5 deg), and an additionalhorizontal bar to complete the horizontal bar(0.5 ¥ 6.6 deg). The luminance of the ILM andTAM figures was 39.4 cd/m2. The fixation andthe square outline were dichoptically presentedto both eyes without binocular disparity. Eachof the transformational motion figures was pre-sented to one eye while continuous flashes werepresented to the other eye.

Procedure. The observers were individuallytested in a dark room and viewed stimuli bymeans of a mirror stereoscope.The viewing dis-tance was 40 cm. The square outlines and fixa-tion crosses were presented throughout theexperiment.

To measure a 75% detection threshold of thecue stimuli in each observer, we controlled theluminance contrast of continuous flashes with aone-up/one-down staircase method, setting theratio of up/down step sizes to 3 (Kaernbach,

Awareness and high-level motion processing 3


1991). The observers pressed the space bar tostart each experimental block and each trialautomatically started after the observers’ pre-vious response.After a blank interval of 300 ms,the presentation of continuous flashes wasstarted. In the ILM condition, 500 ms after thestart of a trial, the ILM cue was presented to theleft or right of the central fixation for 20 ms.Then, the horizontal bar was presented with anexposure duration of 100 ms. The interstimulusinterval between the ILM cue and horizontalbar was kept at 140 ms2. In the TAM condition,500 ms after the start of a trial, the TAM cuewas presented (to the left or right side) for300 ms, and the horizontal bar immediately fol-lowed the cue and lasted for 100 ms. The cuestimuli were presented in half of the trials. Two

hundred milliseconds after the stimulus presen-tation, the observers indicated whether the cuestimulus existed regardless of the location ofthe cue.They reported this by pressing assignedkeys and were informed of the appearance ofthe cue stimuli in advance. Each staircase wasended after 20 reversals of a staircase. Noexplicit feedback for the correctness ofresponses was provided.An experimental blockwhich contained a single staircase was succes-sively repeated twice, and the final five reversalpoints in each of the staircases (10 reversalpoints in total) were averaged to calculate the75% detection threshold for the cue stimuli ofeach of the ILM and TAM conditions in eachobserver.The luminance of each cell in continu-ous flashes was selected from a range betweencertain maximum and minimum luminancevalues. We defined the luminance contrast asthe Michelson contrast calculated on the basisof the maximum and minimum luminance

2These temporal properties were determined so asto obtain optimal motion sensation in each transfor-mational motion based on preliminary observations.

20 ms

500 ms

140 ms

100 ms

300 ms

500 ms

100 ms

ILM stimuli TAM stimuli Continuous flashes

time

Figure 2 Sequences of the illusory line motion (ILM) and transformational apparent motion (TAM) conditionsin Experiment 1. Each of the ILM and TAM stimuli was presented to one eye (the first and second panels),and continuous flashes of colorful, random rectangles were presented to the other eye (the third panel).



values, and in accordance with the observers’responses, these values varied. The luminancecontrast of the mask stimuli increased after acorrect response (i.e., 0.02 log unit of luminancecontrast); correspondingly, the maximum lumi-nance values decreased while the minimumluminance value increased. Furthermore, theluminance contrast decreased after an error(i.e., 0.06 log unit of luminance contrast); corre-spondingly, the maximum luminance valuesincreased while the minimum luminance valuedecreased.

After the determination of the thresholdcontrast of continuous flashes, the observersperformed blocks with motion direction judg-ment and cue detection tasks. The observerspressed the space bar to start each trial andviewed the same stimulus presentation as thatused in the detection threshold measurement.The observers were asked to judge the motiondirection and the presence/absence of a cue ineach trial. The judgments were made using atwo-alternative forced-choice method (“left-ward” or “rightward” for motion direction judg-ment; “present” or “absent” for cue detection)by pressing assigned keys. The luminance con-trast of the mask stimuli was varied at fivelevels from 50% to 400% of a 75% threshold,but fixed within a single block. Each observerperformed 800 trials with four experimentalblocks including two motion conditions (ILMand TAM) ¥ five mask strength conditions (50,100, 200, 300, and 400% of a 75% threshold),and two conditions for cue location (left andright) ¥ 40 replications. In half of the trials, thecue stimuli were not presented. In each block,the trial order was randomized. The order ofthe blocks was counterbalanced across theobservers. It took approximately 2 h for eachobserver to complete the experiment.

ResultsThe mean 75% detection thresholds averagedacross the observers in the ILM and TAM con-ditions were .48 and .15 Michelson contrast,respectively.

For each observer, the visibility of the cuestimuli was calculated as d’ on the basis of thecue detection data obtained from the cue detec-

tion test. Moreover, using data from the cue-present trials, we calculated the proportions oftrials in which the observers reported the illu-sory motion of the bar from the cue side in theILM and TAM conditions. Figure 3a illustratesthe relationship between d’ and the proportionsof illusory motion from the cue side.The regres-sion intercept was considered as an indirectmeasure of ILM and TAM with zero perceptualsensitivity of the cue stimuli (Abrams & Green-wald, 2000; Draine & Greenwald, 1998; Green-wald, Klinger, & Schuh, 1995). We fitted acumulative Gaussian function (Finney, 1971) tothe pooled observational data in each of theILM and TAM conditions. We evaluated thegoodness of fit based on the Kolmogorov-Smirnov test, and confirmed that data were wellfitted by the cumulative Gaussian function(ps > .10). Using the psignifit program imple-mented in MATLAB (Wichmann & Hill,2001a,b), we calculated 95% bootstrap confi-dence intervals by resampling and refittingeach psychometric function 10 000 times. Theestimated regression intercepts at d’ = 0 andtheir 95% confidence intervals are shown inFigure 3b. The results showed that when d’ = 0,both proportions of ILM and TAM were signifi-cantly larger than the chance level, although theproportion of ILM was larger than that ofTAM.

DiscussionThe results showed that both the proportions ofILM and TAM were above chance level whencue stimuli were invisible (d’ = 0), suggestingthat the invisible cue stimuli caused both ILMand TAM. The results in the ILM conditionwere consistent with a previous study (Blanco& Soto, 2009). Moreover, the results in theTAM condition were also consistent with ourprediction. Therefore, the processing of theTAM as well as ILM cues was incompletelysuppressed by continuous flash suppression,even when the visibility of the cues was ulti-mately lowered.

However, it was a critical reservation thatwe did not actually observe the performanceof all of the observers at d’ = 0 in Experiment



1. We merely estimated the performance byfitting a cumulative Gaussian function topooled data: For some of the observers, con-tinuous flashes were not strong enough to

reduce the visibility of the ILM/TAM cuestimuli and thus these cue stimuli were notalways suppressed to a subthreshold detectionlevel. To resolve this issue, in Experiment 2 weemployed stronger masks than those used inExperiment 1. This allowed us to specify lumi-nance contrasts that guaranteed supra- andsubthreshold detection of the cue stimuli foreach observer. In the next experiment, weagain tested the effect of the visibility on theILM and TAM by using the supra- and sub-threshold cue stimuli.

Experiment 2

Method

Observers. Three naive observers (K. Y.,K. S., and Q. K.) and one of the authors (Y. Y.)participated in Experiment 2. They reportedthat they had normal or corrected-to-normalvisual acuity.

Apparatus and stimuli. The apparatus andstimuli used in Experiment 2 were identical tothose used in Experiment 1 except for the fol-lowing: The luminance of each cell of continu-ous flashes was randomly selected from 0 to80.7 cd/m2 and the swap rate was at 10 Hz. TheMichelson contrasts of the masks were 0, .031,.063, .125, .25, .5, .7, .8, .9, and 1. The luminanceof the ILM and TAM stimuli was 42.9 cd/m2;however, only for Y. Y. 43.3 cd/m2 of the lumi-nance was used in the ILM condition to ensuresupraliminal as well as subliminal detectionperformances for the cue stimuli.

Procedure. The observers’ tasks were iden-tical to those in Experiment 1. The trials wereblocked based on one of ten levels of contrastsfrom 0 to 1 and motion type (ILM and TAM).Each block consisted of two conditions for cuelocation (left and right) ¥ 15 replications and 30trials of the no-cue condition. Hence, eachobserver performed 1200 trials in total. In eachblock, the trial order was randomized. Theorder of the blocks was also randomized foreach observer. It took approximately 2 h foreach observer to complete the experiment.

TAM

-1 0 1 2 3 4 5

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Perceptual sensitivity of the cue stimulus (d’)Response p

roport

ion: Illu

sory

motion p

erc

eiv

ed

in the d

irection a

way fro

m the c

ue

ILM

(a)

-1 0 1 2 3 4 5

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Perceptual sensitivity of the cue stimulus (d’)Response p

roport

ion: Illu

sory

motion p

erc

eiv

ed

in the d

irection a

way fro

m the c

ue

(b)

0.5

0.55

0.6

0.65

0.7

ILM TAM

Motion type

Response p

roport

ion: Illu

sory

motion p

erc

eiv

ed

in the d

irection a

way fro

m the c

ue

Figure 3 (a) Relation between the proportions ofillusory motion and perceptual sensitivity of the cuestimulus (d’) in the illusory line motion (ILM) andtransformational apparent motion (TAM) conditionsin Experiment 1. (b) Estimated proportions of illusorymotion when the cue stimulus was invisible (d’ = 0)in the ILM and TAM conditions in Experiment 1.Error bars denote 95% confidence intervals.



ResultsWe calculated the proportions of trials in whichthe observers reported the rightward motion inthe left, right, and no cue conditions for theILM and TAM cues. Figure 4a shows the pro-portion of rightward motion reports as a func-tion of mask contrast.

Moreover, we calculated the proportioncorrect for the detection of the ILM and TAMcues. Figure 4b shows the mean proportioncorrect as a function of mask contrast.We fitteda logistic function to the proportion correctobtained from each observer for the ILM andTAM cues and estimated the 75% detectionthreshold in the mask contrast: .023 (Y. Y.), .02(K. Y.), .03 (K. S.), and .018 (Q. K.) in the ILMcondition and .151 (Y. Y.), .036 (K. Y.), .042(K. S.), and .063 (Q. K.) in the TAM condition.Based on the calculated threshold, we dividedthe data (the proportion of rightward motionreports and proportion correct) into supra- andsubthreshold conditions. A one-sample t-testshowed that the mean proportion of correctdetection in the suprathreshold condition wassignificantly higher than .5 (chance level) in theILM, t(3) = 9.36, p < .003, and TAM conditions,t(3) = 25.98, p < .0002, while that in the sub-threshold condition was not significantly differ-ent from .5 in the ILM, t(3) = 2.79, p > .06, andTAM conditions, t(3) = 1.97, p > .14.

Then, we calculated bias in motion judgmentby subtracting the proportion of rightwardmotion reports in the no-cue condition fromthat in the left and right cue conditions.Figure 4c shows the mean bias in the ILM andTAM conditions. For ILM, a two-way analysisof variance (ANOVA) of bias with visibility(supra- and subthreshold) and cue location (leftand right) as within-participant factors wascarried out. The ANOVA showed a significantmain effect of cue location, F(1, 3) = 381.01,p < .0004, and the significant interaction, F(1,3) = 123.32, p < .003. However, a main effect ofvisibility was not significant, F(1, 3) = 1.86,p > .26. Post hoc tests on the interaction showedsignificant simple main effects of cue location inthe suprathreshold, F(1, 6) = 430.73, p < .0001;Cohen’s d = 5.06, and subthreshold cue condi-tions, F(1, 6) = 10.43, p < .02; Cohen’s d = 1.90.

Moreover, there were significant simple maineffects of visibility in the left, F(1, 6) = 23.19,p < .004; Cohen’s d = 1.84, and right cue condi-tions, F(1, 6) = 50.68, p < .0005; Cohen’sd = 4.73.

For TAM, a two-way ANOVA of bias withvisibility and cue location as within-participantfactors was carried out. The ANOVA showed asignificant main effect of cue location, F(1,3) = 120.38, p < .003, and the significant interac-tion, F(1, 3) = 780.48, p < .0002. However, amain effect of visibility was not significant,F(1, 3) = 1.29, p > .33. Post hoc tests on theinteraction showed significant simple maineffects of cue location in the suprathreshold,F(1, 6) = 336.10, p < .0001; Cohen’s d = 4.85, andsubthreshold cue conditions, F(1, 6) = 7.50,p < .04; Cohen’s d = 1.61. Moreover, there weresignificant simple main effects of visibility in theleft, F(1, 6) = 38.75, p < .0009; Cohen’s d = 4.04,and right cue conditions, F(1, 6) = 15.90,p < .008; Cohen’s d = 1.78.

Furthermore, as in Experiment 1, we calcu-lated the proportions of trials in which theobservers reported the illusory motion of thebar from the cue side in the cue-present trials ofthe ILM and TAM conditions, as a function ofmask contrast. A logistic function was fitted tothe proportion of each observer for the ILMand TAM cues. We calculated 95% bootstrapconfidence intervals by resampling and refittingeach psychometric function 10 000 times. Weestimated regression intercepts at 100% and200% of the threshold mask contrast and their95% confidence intervals in each observer. At200% of the threshold mask contrast, the lowerlimits of 95% confidence interval of both esti-mated proportions of ILM and TAM in allobservers were above .50 (Figure 5). An esti-mated proportion correct at 200% of thresholdmask contrast was .61 � .04 in the ILM condi-tion and .55 � .01 in the TAM condition. One-sample t-tests (two-tailed) revealed that themean proportion correct in each of the ILMand TAM conditions at 200% of thresholdmask contrast was significantly smaller than .75(ps < .04), ensuring that both cues were pre-sented at a subthreshold level. Taken together,the results indicated that ILM and TAM were



0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.2 0.4 0.6 0.8 1

p(R

igh

tward

mo

tio

n)

Mask contrast

Left Right No

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.2 0.4 0.6 0.8 1

p(R

igh

tward

mo

tio

n)

Mask contrast

Left Right No

ILM

TAM

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

YY KY KS QK Mean

Bia

s

Observer

Left

Right -1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

YY KY KS QK Mean

Bia

s

Observer

Left

Right

ILM

TAM

0.5

0.6

0.7

0.8

0.9

1

0 0.2 0.4 0.6 0.8 1

Pro

po

rtio

n c

orr

ect

Mask contrast

0.5

0.6

0.7

0.8

0.9

1

0 0.2 0.4 0.6 0.8 1 P

rop

ort

ion

co

rrect

Mask contrast

(c) Suprathreshold Subthreshold

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

YY KY KS QK Mean

Bia

s

Observer

Left

Right -1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

YY KY KS QK Mean

Bia

s

Observer

Left

Right

(b)(a)

0 0.2 0.4 0.6 0.8 1

Mask contrast

LeftRightNo

LeftRightNo

0.5

0.6

0.7

0.8

0.9

0 0.2 0.4 0.6 0.8 1

Pro

po

rtio

n c

orr

ect

Mask contrast

0.5

0.6

0.7

0.8

0.9

1

Pro

po

rtio

n c

orr

ect

Figure 4 (a) The proportion of rightward motion as a function of mask contrast in the illusory line motion(ILM) and transformational apparent motion (TAM) conditions in Experiment 2. The data of the mean and eachobserver are shown. Error bars denote standard errors of the mean. The vertical arrow indicates estimated75% detection threshold of each observer. (b) The mean proportion correct as a function of mask contrast inthe ILM and TAM conditions in Experiment 2. Error bars denote standard errors of the mean. (c) The meanbias in the supra- and subthreshold conditions of the ILM and TAM conditions in Experiment 2. Error barsdenote standard errors of the mean.



significantly induced by both suprathresholdand subthreshold cues.

DiscussionExperiment 2 replicated the findings of Experi-ment 1. Specifically, the cue stimuli that werepresented at the subthreshold levels could sig-nificantly induce both ILM and TAM. Theseresults again suggest that both ILM and TAMoccur even when the cues were rendered invis-ible by continuous flash suppression.

General discussion

The objective of the present study was toexamine whether the cue visibility that was con-trolled by continuous flash suppression affectedILM and TAM. A previous study showed that aperipheral cue that was rendered invisible byobject-substitution masking could trigger ILM(Blanco & Soto,2009).In contrast, it was unclearwhether invisible cues could induce TAM.Employing continuous flash suppression, thepresent study demonstrated that invisible cuestimuli induced both ILM and TAM when the

cue invisibility was ensured with d’ (Experiment1) and detection threshold (Experiment 2).

The phenomenon observed in the presentstudy is a bit different from the so-far reportedcontextual modulation effect that is induced byinvisible stimuli. Several previous studies havedemonstrated that invisible contextual stimulicaused the change in the appearance of a visiblestimulus. For example, Kawabe and Yamada(2009) showed that invisible peripheral motioncould induce motion contrast in visible centralambiguous motion. Similar kinds of contextualmodulation of visible stimuli by invisiblestimuli have been reported elsewhere (Cai,Zhou, & Chen, 2008; Clifford & Harris, 2005;Kanai, Tsuchiya, & Verstraten, 2006). Thesestudies have focused on a perceptual contrastbetween visible and invisible stimuli on thebasis of spatial integration. In contrast, thepresent study showed that invisible cue stimulialtered the appearance of a visible static bar. Itis thus possible that the effects of invisiblestimuli on the appearance of visible stimuli arenot limited to spatial contrast of orientation ormotion, but are observed in the phenomena

0.5

0.55

0.6

0.65

0.7

0.75

YY KY KS QK

Pro

po

rtio

n i

llu

so

ry m

oti

on

fro

m c

ue

Observer

0.5

0.55

0.6

0.65

0.7

0.75

YY KY KS QK

Pro

po

rtio

n i

llu

so

ry m

oti

on

fro

m c

ue

Observer

ILM TAM

Figure 5 The estimated proportion of illusory motion from cue and its 95% confidence interval (error bars)at a 200% threshold mask contrast for each observer. ILM = illusory line motion; TAM = transformationalapparent motion.



(ILM and TAM) that are not related to spatialcontrast effects, but are mediated by higher-order motion processing.

As a possible explanation of the presentresults, one can argue that ILM and TAM areprocessed in the monocular pathway. It hasbeen suggested that the early component ofvisual processing that occurs monocularly (i.e.,before binocular fusion) is severely suppressed(Blake, Tadin, Sobel, Raissian, & Chong, 2006;Maruya, Watanabe, & Watanabe, 2008) but notabolished by an interocular mask (Cai et al.,2008). Because in the present study ILM andTAM were observed even when the visibility ofcue stimuli was suppressed by continuous flashsuppression, which was a kind of the interocularmasking, it was possible that the monocularpathway at least partially contributes to theseillusory motion percepts. However, Hikosakaet al. (1996) have observed that ILM occurredeven when a cue and a bar stimulus were pre-sented to different eyes, indicating that at leastILM is processed in the binocular pathway thatis susceptible to interocular masking. More-over, it is difficult to clearly say that a phenom-enon that is suppressed by continuous flashsuppression is always processed in the monocu-lar pathway, because, although it is equivocalwhether the visual processing for emotionalstimuli is monocular or binocular, the emo-tional stimuli that are rendered invisible bycontinuous flash suppression can still affect thevisual orientation discrimination (Jiang, Cos-tello, Fang, Huang, & He, 2006) or time percep-tion of visual stimuli (Yamada & Kawabe,2011). Thus, it is hard to conclude as to whetherILM and TAM are processed in the monocularor binocular pathway at this stage.

The suppression of visual transients by con-tinuous flash suppression may provide a morerigorous explanation for the present study. InMotoyoshi and Hayakawa (2010), an invisibleorientation annulus without visual transientscaused the orientation contrast in a centralvisible orientation patch. That is, even thoughthe visual signals without the transients do notenter the visual awareness, they still effectivelyinfluence on-going visible events in the spatialvicinity. In the present study it is possible that

continuous flash suppression effectivelyreduced the transients of neural signals sent tohigher-order motion processing. Hence, theneural signals could not cause the visual aware-ness for the cue stimuli. However, they wereenough to stimulate higher-order motion pro-cessing. We suggest that this is why the invisiblecue stimuli caused ILM and TAM.

A more reliable, precise measurement ofsubjective invisibility may be required in assess-ing the influence of invisible on visible infor-mation. Although the present results demon-strated that invisible cues induced ILM andTAM, it was unclear that the cue stimuli werecompletely invisible. We used an indirectmeasure of visibility, that is, an estimated inter-cept at a null sensitivity point (or d� = 0) thatwas extrapolated by a regression technique(Abrams & Greenwald, 2000; Draine & Green-wald, 1998; Greenwald et al., 1995). Based onthis technique, we inferred that perceptual sen-sitivity for the cue stimuli was nonlinearlyrelated to mask contrast. However, a true rela-tion between them around a null sensitivitypoint is unknown (Hannula, Simons, & Cohen,2005). Moreover, in Experiment 2, we set a 75%correct response as a threshold for the detec-tion of cue stimuli, and weak but significantillusory motion percepts were obtained in thesubthreshold condition (i.e., at a mask contrastof 200% threshold). In contrast, it is still pos-sible to argue that the weak illusory perceptsmay have stemmed from the data from slighttrials in which the cues were accidentallyvisible. Thus, it is precise to conclude that whatthe present results demonstrated was that thesubthreshold (but sometimes visible) cues con-tributed to the generation of illusory motionpercepts. In light of these issues, it is necessaryto devise the measurement of cue invisibility toprecisely investigate the relation betweenvisual transients, visual awareness, and motionprocessing.

References

Abrams, R. L., & Greenwald, A. G. (2000). Parts out-weigh the whole (word) in unconscious analysisof meaning. Psychological Science, 11, 118–124.



Bavelier, D., Schneider, K., & Monacelli, T. (2002).Reflexive gaze orienting induces the line motionillusion. Vision Research, 42, 2817–2827.

Blake, R., Ahström, U., & Alais, D. (2000). Perceptualpriming by invisible motion. PsychologicalScience, 10, 145–150.

Blake, R., Tadin, D., Sobel, K. V., Raissian, T. A., &Chong, S. C. (2006). Strength of early visualadaptation depends on visual awareness. Pro-ceedings of the National Academy of Sciences ofthe United States of America, 103, 4783–4788.

Blanco, M. J., & Soto, D. (2009). Unconscious percep-tion of a flash can trigger line motion illusion.Experimental Brain Research, 192, 605–613.

Brainard, D. H. (1997). The Psychophysics Toolbox.Spatial Vision, 10, 433–436.

Cai, Y., Zhou, T., & Chen, L. (2008). Effects of bin-ocular suppression on surround suppression.Journal of Vision, 8 (9), http://journalofvision.org/8/9/9/.

Chica, A. B., Charras, P., & Lupiáñez, J. (2008). Endog-enous attention and illusory line motion dependon task set. Vision Research, 48, 2251–2259.

Clifford, C. W. G., & Harris, J. A. (2005). Contextualmodulation outside awareness. Current Biology,15, 574–578.

Draine, S. C., & Greenwald, A. G. (1998). Replicableunconscious semantic priming. Journal ofExperimental Psychology: General, 127, 286–303.

Field, D. J., Hayes, A., & Hess, R. F. (1993). Contourintegration by the human visual system: Evi-dence for a local “association field”. VisionResearch, 33, 173–193.

Finney, D. K. (1971). Probit analysis. Cambridge:Cambridge, University Press.

Fracasso, A., Caramazza, A., & Melcher, D. (2010).Continuous perception of motion and shapeacross saccadic eye movements. Journal ofVision, 10 (13), http://www.journalofvision.org/content/10/13/14.

Ghorashi, S., Jefferies, L. N., Kawahara, J., &Watanabe, K. (2008). Does attention accompanythe conscious awareness of both location andidentity of an object? Psyche, 14, http://theassc.org/vol_14_2008.

Green, D., & Swets, J. (1966). Signal detection theoryand psychophysics. New York: Wiley.

Greenwald, A. G., Klinger, M. R., & Schuh, E. S.(1995). Activation by marginally perceptible(“subliminal”) stimuli: Dissociation of uncon-scious from conscious cognition. Journal ofExperimental Psychology: General, 124, 22–42.

von Grünau, M., Dubé, S., & Kwas, M. (1996). Twocontributions to motion induction: A preatten-tive effect and facilitation due to attentionalcapture. Vision Research, 36, 2447–2457.

Hannula, D., Simons, D. J., & Cohen, N. (2005).Imaging implicit perception: Promise and pit-falls. Nature Reviews Neuroscience, 6, 247–255.

Hikosaka, O., Miyauchi, S., & Shimojo, S. (1993a).Focal visual attention produces illusory temporalorder and motion sensation. Vision Research, 33,1219–1240.

Hikosaka, O., Miyauchi, S., & Shimojo, S. (1993b).Voluntary and stimulus-induced attentiondetected as motion sensation. Perception, 22,517–526.

Hikosaka, O., Miyauchi, S., & Shimojo, S. (1996). Ori-enting of spatial attention: Its reflexive, compen-satory, and voluntary mechanisms. CognitiveBrain Research, 5, 1–9.

Hoffman, D. D., & Richards, W. A. (1984). Parts ofrecognition. Cognition, 18, 65–96.

Ishigami, Y., Klein, R. M., & Christie, J. (2009).Exploring the modulation of attentional captureby attentional control settings using perfor-mance and illusory line motion. Visual Cognition,17, 431–456.

Jiang, Y., Costello, P., Fang, F., Huang, M., & He, S.(2006). A gender- and sexual orientation-dependent spatial attentional effect of invisibleimages. Proceedings of the National Academy ofSciences of the United States of America, 103,17048–17052.

Kaernbach, C. (1991). Simple adaptive testing withthe weighted up-down method. Perception andPsychophysics, 49, 227–229.

Kanai, R., Tsuchiya, N., & Verstraten, F. (2006). Thescope and limits of top-down attention in uncon-scious visual processing. Current Biology, 16,2332–2336.

Kawabe, T., & Yamada, Y. (2009). Invisible motioncontributes to simultaneous motion contrast.Consciousness and Cognition, 18, 168–175.

Kawahara, J. (2002). Facilitation of local informationprocessing in the attentional blink as indexed byshooting line illusion. Psychological Research, 66,116–123.

Kawahara, J., & Yokosawa, K. (2001). Preattentiveperception of multiple illusory line motion: Aformal model of parallel independent-detectionin visual search. Journal of General Psychology,128, 357–383.

Kawahara, J., Yokosawa, K., Nishida, S., & Sato, T.(1996). Illusory line motion in visual search:Attentional facilitation or apparent motion? Per-ception, 25, 901–921.

Maruya, K., Watanabe, H., & Watanabe, M. (2008).Adaptation to invisible motion results in low-level but not high-level aftereffects. Journalof Vision, 8 (11), http://journalofvision.org/8/11/7/.



Motoyoshi, I., & Hayakawa, S. (2010). Adaptation-induced blindness to sluggish stimuli. Journalof Vision, 10 (2), http://journalofvision.org/10/2/16/.

Pelli, D. G. (1997). The VideoToolbox software forvisual psychophysics:Transforming numbers intomovies. Spatial Vision, 10, 437–442.

Shimojo, S., Miyauchi, S., & Hikosaka, O. (1997).Visual motion sensation yielded by non-visuallydriven attention. Vision Research, 37, 1575–1580.

Singh, M., Seyranian, G. D., & Hoffman, D. D. (1999).Parsing silhouettes: The short-cut rule. Percep-tion and Psychophysics, 61, 636–660.

Tse,P.U. (2006).Neural correlates of transformationalapparent motion. Neuroimage, 31, 766–773.

Tse, P. U., & Caplovitz, G. P. (2006). Contour discon-tinuities subserve two types of form analysis thatunderlie motion processing. Progress in BrainResearch, 154, 271–292.

Tse, P. U., Cavanagh, P., & Nakayama, K. (1998). Therole of parsing in high-level motion processing.In T. Watanabe (Ed.), High-level motion process-ing: Computational, neurobiological, and psycho-physical perspectives. Cambridge, MA: MITPress, pp. 249–266.

Tse, P. U., & Logothetis, N. K. (2002). The duration of3-D form analysis in transformational apparentmotion. Perception and Psychophysics, 64, 244–265.

Tsuchiya, N., & Koch, C. (2005). Continuous flashsuppression reduces negative afterimages.Nature Neuroscience, 8, 1096–1101.

Wichmann, F. A., & Hill, N. J. (2001a). The psycho-metric function. I. Fitting, sampling, and goodnessof fit. Perception and Psychophysics, 63, 1293–1313.

Wichmann, F. A., & Hill, N. J. (2001b). The psycho-metric function. II. Bootstrap-based confidenceintervals and sampling. Perception and Psycho-physics, 63, 1314–1329.

Yamada, Y., & Kawabe, T. (2011). Emotion colorstime perception unconsciously. Consciousnessand Cognition, 20, 1835–1841.

Yamada, Y., Kawabe, T., & Miura, K. (2008). Mislo-calization of a target toward subjective contours:Attentional modulation of location signals. Psy-chological Research, 72, 273–280.

(Received December 8, 2010; accepted September 6,2011)



illusory line motion and transformational apparent motion during continuous flash suppression

Documents